fix(tracing): flush guards on Ctrl-C and doctor-subcommand exit

[bug-ops]

Summary

• std::process::exit runs no destructors, so early_ctrlc's Ctrl-C handler and the Doctor/Gonka::Doctor/Cocoon::Doctor subcommand exits (src/runner.rs) bypassed TracingGuards::drop's take-once-cell flush logic entirely, truncating the log file and/or local Chrome trace on every early exit through those paths. Unlike the SIGTERM race window fixed in log_guard (rolling file WorkerGuard) has same SIGTERM-flush gap as chrome_guard had #6693, the Ctrl-C gap was a 100%-reproducible bypass on every platform.
• Added exit_with_flush and flush_and_exit_on_ctrlc in src/tracing_init.rs, and extracted the previously-duplicated lock-hold-across-flush logic into a shared take_and_flush_guard_cells helper reused by TracingGuards::drop, spawn_sigterm_flush_task, and the new Ctrl-C path.
• tokio::signal::ctrl_c() is cross-platform, so this also closes the non-Unix half of the gap that log_guard (rolling file WorkerGuard) has same SIGTERM-flush gap as chrome_guard had #6693 left open.

Closes #6696

Scope notes

Two sites sharing the same std::process::exit pattern (runner.rs:1186 Notify::Test, :1231 UrlScheme::Status) and 5 deep-link-feature-gated sites inside handle_url_open share the same bug class but were intentionally left out of scope — a follow-up issue will be filed for them.

Review chain

• Debugger diagnosed the exact root cause and fix shape (reuse the spawn_sigterm_flush_task clone-and-race pattern).
• impl-critic's first pass returned significant: the DRY extraction had silently inverted TracingGuards::drop's flush order (chrome-then-log → log-then-chrome) and left 3 comments stale. Developer confirmed via git archaeology on Local trace files truncated when zeph is killed via pkill (FlushGuard never dropped) #6683/log_guard (rolling file WorkerGuard) has same SIGTERM-flush gap as chrome_guard had #6693 that the original order was incidental (not load-bearing — the two writers are independent files), kept the new order, and corrected the 3 comments. Re-verified approved on second pass.
• Code reviewer requested a CHANGELOG.md entry (matching the Local trace files truncated when zeph is killed via pkill (FlushGuard never dropped) #6683/log_guard (rolling file WorkerGuard) has same SIGTERM-flush gap as chrome_guard had #6693/fix(tracing): flush log_guard on SIGTERM alongside chrome_guard #6695 entries in the same chain) and a correction to the now-stale "remains truncating on both Unix and non-Unix" sentence in the existing log_guard (rolling file WorkerGuard) has same SIGTERM-flush gap as chrome_guard had #6693 entry. Both addressed, re-reviewed approved.

Test plan

• cargo +nightly fmt --check
• cargo clippy --profile ci --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,testing" -- -D warnings
• cargo nextest run --config-file .github/nextest.toml --workspace --features "desktop,ide,server,chat,pdf,scheduler" --lib --bins (15258/15258 passed, including all tracing_init tests unmodified)
• RUSTFLAGS="-D warnings" RUSTDOCFLAGS="--deny rustdoc::broken_intra_doc_links" cargo doc --no-deps --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• gitleaks protect --staged --no-banner --redact
• No new automated test was added for exit_with_flush/flush_and_exit_on_ctrlc themselves — both call std::process::exit and are untestable in-process (verified no assert_cmd/subprocess-exit test harness precedent exists in the repo). take_and_flush_guard_cells is indirectly covered via existing TracingGuards::drop tests.